Assessing country UHC performance relative to spending€¦ · Health Financing for UHC, Barcelona, Spain 8-12 June 2015 Assessing country UHC performance relative to spending: a

Health Financing for UHC, Barcelona, Spain 8-12 June 2015

Assessing country UHC performance relative to spending:

a DEA investigationMatthew Jowett (WHO Geneva)Maria Petro Brunal (Global Fund)

Gabriela Flores (WHO Geneva)Jonathan Cylus (European Observatory/WHO)

29th November 2019Geneva, Switzerland

Health systems’ efficiency’ Special Interest Group

https://www.healtheconomics.org/page/HSE


Motivation for the analysis•Growing dissatisfaction with use of health expenditure

targets in the health financing dialogue•Based on a wide range of assumptions e.g. scope of

services, extent of scaling up. timeline. Total or public? Evidence base often weak.• Some value for advocacy, but can have negative effect on

direction of country level discussions, action. •Risks sending message that progress is not possible

without significant increases in spending.•WHO publishes health expenditure data (GHED), and

health coverage data (GHO), typically considered separately.•We analyse variation in performance relative to spending,

in 83 low and middle income countries (2012 data)


Numerous targets exist

$12-22 per capita

1993 Abuja: 15% GGE to health

2001$60 per capita

2009

<20% OOP/THE 5%GDP

2010$44-$80

per capita

2010 $86 per capita/ 5% GDP

2014

Health expenditure estimates


Our decisions• A technique to assess the

relationship between the level of public expenditure on health in a country and its performance on agreed UHC tracer indicators.

• A method flexible enough to accommodate one input and multiple outputs simultaneously.

Our requirements• We used non-parametric DEA; linear

programming method which allows us to determine how well an entity, in our case a country, converts an input (health spending) into outputs (performance).

• We assigned the constraint that all outputs be given a weight greater than 0, to ensure all were evaluated, whilst permitting flexibility in the assignment of weights. Without these constraints, poor performance on some indicators may be ignored.

• We used an output-oriented model, which reports the extent to which a unit can be expected to increase its outputs at its given level of inputs.

• We conducted three sensitivity analyses.


Data envelopment analysis:

in algebraic terms


Single input: what governments spend

•We use public rather than total health expenditure, given the widespread evidence that private sources (voluntary insurance, OOPs) contribute little to UHC progress at the system level.

•Public spending on health as % GDP gives very little variation across our sample.

•Indicator used: public spending per capita on health (PPP$ for 2012) – from WHO Global Health Expenditure Database.


Six output measures of UHC performanceTB treatment (% coverage) DPT among 1-year-olds (%

coverage)

Live births attended by skilled personnel (% cov.)

Family planning (% coverage)

Antiretroviral therapy (% cov) Total public spending (GGHE) as a % total health spending

Sample: 83 low and middle income countries (>1.5m pop)Source: GHED & GHO for 2012 or nearest

- 64/83 countries achieve 80%- Of those achieving, lowest spenders are in region of $8-20 per capita

- 65/83 countries achieve 80%- Of those achieving, lowest spenders around $8-30 per capita

- 43/83 countries achieve 80%- Of those achieving, lowest spenders $41-49 per capita, but average is higher = $450 pc

- 12/83 countries achieve 80%- Of those achieving, the lowest spender is $192 per capita

- 0/83 countries achieve 80%- Best performers spend little:1st = 67% coverage for $41 pc2nd = 66% coverage for $92 pc

>80% in 2/83 countries(spending $1642 Cuba & $788 Romania).

>70% in 11/83 countriesProxy

DEA scores

Assesses performance (outputs) relative to inputs in countries with similar levels of public spending per capita on health. Each country is given a score between 0-100%. A score of 100% is given to the “best performer” for

a given level of spending. A score of 50%, for example, implies a performance 50% that of best performer.

Myanmar Cambodia Malawi

Rwanda Thailand Cuba


Standard deviation of DEA scores by quintile

QuintileLevel of public spending on

health per capita

Number

countriesStd. dev.

Mean DEA

scoreMin DEA score

1 GGHE less than $37.50 17 14.4% 77.0% 45.5%

2GGHE between $38.20 and

$59.6017 9.9% 83.0% 59.2%

3GGHE between $60.90 and

$198.1016 10.7% 86.8% 57.1%

4GGHE between $207.60

and $500.3017 6.5% 88.1% 79.1%

5 GGHE higher than $520.50 16 5.6% 87.8% 78.3%

Disaggregated performance by spending quintiles

Service coverage: - systematic increase in performance with increased public spending

- systematic fall in variation across countries (less poor performers)

Financial protection: - performance increases in Q4 and Q5. High variation remains

N=83, n=16/17

<$38 $38 to $60 $61 to $198 $208 to $500 >$520


Sensitivity analysisalternative financial protection proxy (1-OOP/THE)

Initial proxy excludes VHI, external resources flowing outside

government channels, and OOPs

Five of six countries remain best performers (Cambodia dropped)


Sensitivity analysis

• Sensitivity #2: average values for 2008-2012 to smooth transitory shocks. Best performers remain the same except Rwanda replaced by China

• Sensitivity #3: Removal of imputed data (23 countries had at least one value imputed). Remaining sample = 60. Myanmar and Rwanda dropped as best performers.

• Re-run of model with updated data (2014) where available: only 57 countries. Cambodia, Thailand and Cuba remain best performers. Myanmar & Rwanda dropped due to lack of data. Malawi drops. Madagascar & El Salvador added.

• Limitations of the study• financial protection proxy: will re-run model with improved financial

protection data (some descriptive preliminary results)• Not a 1:1 relationship between inputs and outputs


Not one country in our sample achieves both 80% coverage on each

of the five service indicators and >70% public health spending as

%THE. On ART coverage, the best performers are some of the lowest

spenders. Targets are of limited value.

Observe large variation in UHC performance for the money spent at

low levels of public spending.

Suggests many countries can improve spending, even at very low levels e.g. <40$ per capita; however, non-health

system factors also important.

ABSOLUTE LEVELS OF PUBLIC SPENDING MATTERS

We observe systematic improvement in service coverage as public

spending increases, as well as convergence across countries.

Whilst financial protection also improves with public spending

(especially >$200), progress remains highly variable across countries, even

at public spending >$520 pc.Improving financial protection requires development of risk-sharing institutions and related policies. Private spending?

OBSERVATIONS & POLICY MESSAGES

Health Financing for UHC, Barcelona, Spain 8-12 June 201529/11/2019

| Financial protection in health

14

• Source: https://www.who.int/healthinfo/universal_health_coverage/report/2019/en/

• Data available from http://apps.who.int/gho/portal/uhc-financial-protection-v3.jsp

Updates and possible future work• SDG indicators to track progress

towards UHC within the SDGs have been adopted (listed next slide)

• Comprehensiveness of SDG UHC indicators has improved

• more countries (and more countries with panel data)

• Possible to explore countries trajectories and how that relates to public/private levels of spending (methods to be defined)

• VERY preliminary descriptive analysis using cross-sectional SDG indicators suggests most of the findings still hold (see next slides)

https://www.who.int/healthinfo/universal_health_coverage/report/2019/en/

http://apps.who.int/gho/portal/uhc-financial-protection-v3.jsp


Disaggregated performance using SDG indicators of UHC by spending quintiles

• Different sub-sample: N=71, n=15/14

Service coverage (SDG 381):

- Linear increase in performance with increased public spending up to Q4

- Significant variation observed across countries at all quintiles

Financial protection (SDG 382, 10%):

- No clear pattern observed


Thank you

www.who.int/health_financing/documents/no-magic-number


• Composite index based on geometric means of 4 dimensions, each one based on geometric means of tracer indicators

• Presented on a 0-100 scale with “high score indicating better performance”

• Limitation: not a measure of population coverage

SDG UHC indicators – 381 service coverage




Proportion of the population with household out-of-pocket spending on healthgreater than 10% or 25% of household’s total consumption or income

2015

SDG UHC indicators – 382 catastrophic health spending

• Impact measure of financial hardship at population level

• Reflects ethical concern: no one should have to choose between spending on health and spending on other basic goods and services (e.g. food, education)

• Limitations:

• Insensitive to financial barriers to access (unmet need is not captured)

• SDG definition tends to be more concentrated among the rich than the poor


Countries are at different stages in service coverage and financial protection as tracked by SDG indicators 381 & 382

125 countries Sub-sample 71 countries

• N=71 countries• Latest estimate available for 382• 381 estimate matched to year of 382 estimate

(2008-2015)

• Only 3 good performers in both dimensions

• Only 10 good performers in 381 but poor performance in 382(large variations)

Assessing country UHC performance relative to spending€¦ · Health Financing for UHC, Barcelona, Spain 8-12 June 2015 Assessing country UHC performance relative to spending: a

Documents